Chapter 9
Timecode and linear postproduction

The transfer suite

After rushes have been viewed to sort the wheat from the chaff, the 'wanted' takes should be identified and transferred to a format suitable for the off-line edit. Traditionally, Hi-band U-matic machines have been used for this, as the format supports timecode. Some post-production houses, at least one broadcaster, and some manufacturers, have developed VITC generators that will lay code onto Lo-band and VHS. Laying down a proper time-address signal is preferable to burning-in the code, as the off-line editor has the options of switching it in or out, whether to put it in a 'letter-box' or superimpose it, and - important where to place it on the screen. If code has to be burnt in, consideration should be given as to where to place it within the image. Some editors will wish to see the bottom of the picture, as footsteps can be used to help sync pictures with sound. Superimposed code can be more difficult to read, but is less obtrusive than code in a letter-box. Whatever method is used, the time addresses must correspond with those on the rushes. If the off-line tapes are from film transfer, both KeyKode® and the in-camera recorded code should be transferred, and the time addresses also stored in a computer database for later conforming and negative cutting once the offline edit has been done.

Whatever method is used to marry code with the off-line copy, it should always be transferred via a regenerator, which should be set to 'external' and 'auto'. If possible, both VITC and LTC should be transferred to the off-line copy. It is important that they should be identical. If they are not, the addresses could unexpectedly change during the off-line edit as the play VCR switches automatically between the two.

Off-line editing

It can be cost-effective to make the major edit decisions off-line, away from the expensive (both in time and personnel terms) edit suite that will be used to complete the final master edited tape. The tape format, edit controller and other hardware used for off-line editing will be less expensive, and the off-line editing may be personally undertaken by the director. At its very simplest the off-line edit may be a 'paper' edit, where the director sits down with a VHS copy of the road tapes, with timecode 'burned in' within a letter-box-shaped window in the picture (a 'window' dub), and writes down the 'in' and 'out' times by hand (Figure 9.1). The decisions may be stored in a timecode organizer or, in more sophisticated systems, on floppy or hard disc by means of a controlling microcomputer, for later transfer to the on-line edit controller.

Figure 9.1 In its simplest form an EDL can be handwritten.

The two most common off-line edit methods will obtain a 'rough cut' of the final version by means of assembly or insert editing. It may be that little thought will be given to the problems of the 8-field sequence when editing composite pictures, the edit controller being switched to 2- or 4-field. Some off-line editors may even work from the control track. In both cases, the matter of the correct sequence will be attended to during conforming, prior to the on-line edit. For this reason any audio postproduction should be done only after conforming.

Assembly editing

Figure 9.2 In assembly editing (a) programme, control and timecode tracks are laid down together consecutively as the edit proceeds. Insert edits (b) involve striping the tape with control track prior to the edit. Video and audio can be recorded individually. When striping the tape, the opportunity is often taken to lay LTC as a stripe of contiguous time addresses.

Assembly editing consists of adding successive video, audio, control track (in the case of video and some digital audio formats) and timecode onto a tape which is effectively virgin. The new information is butt-joined to that already on tape, using the control track, and maybe timecode, to ensure a seamless join (Figure 9.2a). (In this respect it should be noted that assembly editing on Prodigi and some implementations of the R-DAT format require pre-striping.) The video, audio, control and timecode tracks are extended as the process continues. Before each edit the tape is backwound to a point before the end of the previous sequence, and on the approach to the edit point the previously-laid control track is used to guide the servo systems to ensure correct continuation of synchronization. If the edit controller can generate a list of both original and corresponding off-line edit master time addresses during this process, the record machine can be set up for record-run code. This will result in a continuous and contiguous code on tape, with the corresponding original addresses on disc. This will permit later insert-editing of the tape, should this prove necessary. If this degree of sophistication is not available, the recording machine should be set to 'external' and 'regenerate'. The time addresses recorded will be those of the working copies of the original tapes. If VITC is being recorded, the lines should be the same as those on the working copies of the road tapes. If this is not done there is the possibility of a VITC reader receiving different information from different lines, if it is set to auto. If an edit list is stored on disc, it should be printed out in hard copy each time it is modified, to protect against loss in the event of a system crash.

During assembly editing, both video and sound must be laid down together; in particular, audio alone cannot be laid down because of the need for a control track. Note that later insert-editing on the tape can be performed only if continuous control track is present without any disturbance at the edit point, together with a timecode address track that is both continuous and contiguous.

Insert editing

Insert editing permits video, audio and sometimes timecode to be laid down individually. The process requires the tape to be pre-striped with both control track and longitudinal timecode (LTC). This is usually done by pre-recording 'picture black' or 'black and burst' (Figure 9.2b), though some editors prefer a coloured video field. Reference to the colour subcarrier will be necessary for sophisticated editing in a composite environment, to prevent picture disturbance at the edit point. Component videotape recorders do not, in themselves, require reference to colour subcarrier. However, a line of colour sub-carrier is recorded during the vertical interval when recording from a composite source.

VITC cannot be pre-striped as it will be overwritten when video information is recorded. It is important that longitudinal timecode is not overwritten while insert-editing, otherwise any subsequent editing or audio post-production will be made very difficult. It is sensible to generate VITC from the pre-striped LTC during the video insert edit as both timecodes will be present and identical. The recorder's internal regenerator should be set to 'internal' and 'regenerate'. The videotape machines employed later in the post-production process will then have the choice of both timecode forms, for maximum flexibility.

Pre-roll requirements

Since all machines require a few seconds of pre-roll in order to synchronize before performing an edit, it is important that the timecode addresses over this period are contiguous. If they are not, the edit controllers/synchronizers may pick up the discrepancy and abort the edit. The pre-striped timecode and picture black must be synchronous (and colour-framed when working in composite). If not, lack of synchronism between timecode and video frames will cause edits to abort, and lack of correct colour framing in composite will cause problems with sophisticated edits. If the recorder has an inbuilt timecode generator this will need to be set to 'internal'. If a separate regenerator is used, it must be fed with the same syncs as the video machine by way of reference. It is not a good idea to momentarily jam-sync a timecode regenerator, then remove its reference. It may well not be stable enough to maintain accurate synchronization for more than a few minutes.

If timecode on the edit master has lost synchronism with the video, it may be possible to retrieve the situation by replaying the tape from the head with the internal timecode generator set to regenerate, then switching the LTC track into record. This will ensure that the regenerated code will remain in sync with the video. If an external timecode generator is used the same procedures apply, but it must be fed with reference video from the videotape recorder, not station syncs.

When insert-editing, conventional practice is to have the edited programme start at lOh 00m 00s 00f. It is not a good idea to start the programme at '00' hours because of the 'midnight crossing' problem. One way of doing this is by laying down timecode from the head of the tape starting at 09h 57m 00s 00f. This will allow the recording a leader at the head of the tape.

The edit decision list

In its basic form the edit decision list will be a written list containing reel numbers, in and out times of the edits with corresponding start and finish times of the source tapes and details of the individual edits, such as video only, cut, wipe etc. (Figure 9.3). At its most sophisticated, the EDL will also contain details such as colour grading of the source tapes, details of split edits etc., with this information included in the EDL before autoconforming to permit most efficient use of time. Whatever its sophistication, compilation of the EDL will rely on timecode. Both on- and off-line editing sessions are greatly helped by having accurate production logging (shot lists) with timecode.

Editing and the colour frame sequence

As discussed in Chapter 1, the relationship between colour sub-carrier and video syncs varies over an 8-field sequence in PAL, 4-field in NTSC and SECAM. For this reason, when editing together material from a variety of composite sources, the sub-carriers on either side of the edit must be brought into phase. If this is not done there may be picture disturbance following the phase discontinuity, when the edited material is replayed. This is caused by the sub-carrier regenerator in the receiver or picture monitor changing frequency and/or phase in order to resync. To prevent this happening, the timebase corrector(s) (TBCs) associated with the source tape machine(s), in combination with the edit controller, will time-shift the picture information to ensure the correct phase relationship exists both sides of the edit.

Figure 9.3 A computer-generated EDL showing menu (1), system messages (2), and the actual EDL (3) with information concerning the individual edits.

When undertaking simple editing, i.e. cutting or mixing between dissimilar programme material, the TBC corrects for phase discontinuities by time-shifting the video signal (in theory in either direction, but in practice inserting a delay), usually horizontally, by multiples of one quarter-cycle with respect to the horizontal sync pulses, and / or down by a line to obtain a half-cycle phase shift, to bring the sub-carriers in phase across the edit. On cuts or mixes between dissimilar material this horizontal or vertical shift is not going to be noticeable unless the monitor is underscanned. Certainly, without any common spatial reference point within the picture, the small shift in the position of the programme material is irrelevant.

The situation is different when cutting or mixing between identical picture material, for example during an animation sequence with a common background, or when shortening or lengthening a scene, a process known as 'invisible' editing. Here, any timing shift will be noticeable as horizontal and/or vertical picture displacement. The way of dealing with this is to instruct the edit controller to perform a colour-framed edit. This is done by amending the IN point(s) of the source machine(s) so that the sub-carrier(s) when extrapolated across the edit will be in phase with the sub-carrier on the edit master.

The correct phase relationship cannot be obtained by decoding the composite signal into its components, then re-encoding with correctly phased sub-carrier, because decoded composite video carries a residual colour sub-carrier, which cannot be filtered out without degrading the signal (colour sub-carrier lies within the luminance bandwidth). This residual sub-carrier, carried across a non-colour-framed edit, will interfere with the luminance signal, causing various degrees of cancellation. Final phasing may have to be performed manually if the S-C/H phase relationship is not close enough (+ 20° is the recommended tolerance). Material acquired and edited totally within a component environment will not have these constraints.

Whatever technique is used to match sub-carrier phases, the edit controller and TBC need to know the phase of all sub-carriers so that the correct time shift can be inserted. This is done most conveniently if there exists a fixed and standard relationship between timecode addresses and sub-carrier to horizontal phase because the timecode(s) on source and record machines can be examined to establish the various phase relationship(s). The fixed and standard relationship between timecode and colour sub-carrier is indicated by the presence of the colour frame flag. If this flag is present in all source tapes and the pre-striped record master tape, the phase relationship can be established. It is for this reason more than any other that no composite video source should ever be fed with timecode that it is not referenced to a colour sync pulse generator. All sync pulse generators intended for use in composite editing suites, or component suites where composite sources may be handled, must generate a correctly colour-framed sequence of pulses and colour burst. Not all SPGs maintain the correct sub-carrier-to-sync relationship. If relaying damaged timecode on a road tape, the timecode regenerator must be fed with the off-tape video as reference.

The relationship between timecode and the 8- or 4-field sequence will be correct only if the sub-carrier phase is correct with respect to the horizontal sync pulse edges (S-C/H phase). Timebase correctors associated with videotape machines have the ability to vary S-C/H phase incrementally. The phase should be correct within 10° for trouble-free editing, and should always be checked for all source machines to avoid ambiguities. It should be suspect if problems arise when a colour-framed edit fails. A timecode sync monitor can display information about colour framing on a picture monitor (see Figure 3.19, p. 57).

Audio post-production

As audio signals are continuous, it is perfectly possible to slip the audio signal within each timecode frame, something not possible with video, to aid synchronization. With 80 bits per timecode word (frame), offsets are possible down to 1/80 or less of an equivalent video frame, an incremental time of 1 / 2000 s at 25 fps. One sound station on the market can slip individual channels relative to each other and the master VCR or synchronizer timecode by as little as 1 /10 000 s. Analogue and audiotape recorders do not have timebase correctors, so it is important that any control device is able to smooth out any timing jitter on incoming timecode if wow and flutter are to be avoided.

Synchronizers

Synchronizers control all transport functions of the machines, including 'stop', 'play', 'fast forward', 'rewind' and 'speed'; they enable a number of machines to play together in synchronization; they allow timecode offsets between machines both in integer frames and fractions of a frame. They may emulate the characteristics of a VCR in order to allow an audio machine to be controlled from an edit controller. To do this efficiently each device will require details of its machine's ballistics (acceleration, deceleration, spooling speed, variable speed play range, nature of tachometer pulses). Commonly a synchronizer can be programmed either by software or by hardware changes such as resistor links, to 'recognize' the tape machine associated with it. Some synchronizers have the ability to 'learn' a machine's ballistics, and some, although pre-programmed, can still confirm the ballistics on switch-on.

Masters and slaves

As a rule, one synchronizer is designated 'master'. This is normally the one controlling the videotape recorder in video/audio post-production. This is because it is more difficult to make a videotape recorder slave to an external source than it is an audiotape recorder. Designating more than one synchronizer as master causes total confusion, as each of the synchronizers concerned tries to control the others. With some audio formats this can cause physical damage. If one synchronizer is designated 'master', the other synchronizers control the slave machines, which are thereby forced to run in synchronism with the master. It is important that the synchronizers controlling these machines do not lock them too firmly to the master, as any speed variation in the master would then be reflected in the slave machines. It is also important that the slave synchronizers 'roll-over' any short breaks in timecode which may be caused, for example, by dropout on the master, if the edit is not to crash. All synchronizers should have a selection of locking ranges available, and a selection of timecode dropout limits. Some may have the ability to reference to incoming timecode on machine start-up, switching to incoming reference syncs once the slave machine is locked. One point to remember is that when a tape machine stops, its parked position will depend on its braking ballistics; so it may well riot be parked in synchronism with the master. If this is the case, it will have to search for synchronism on start-up. This will take some time, depending on its ability to play over a range of speeds; the smaller the range, the longer the lock-up time.

Chase synchronizers

Chase synchronization is the simplest form of control. Under this system the slave machine follows the master. The slave cannot park until the master has parked. Its speed follows that of the master, within limits. The slave will lock to timecode supplied to it from the master. Should the master shuttle at a faster rate than the slave is able to follow, the slave's capstan drive will disengage, the tape will lift off the heads, and the slave machine will go into spool. This may result in the slave running out of control (a 'speed hit'). The consequence is a possible run-off at the end of tape. More likely, the result will be excessively long synchronization times. To prevent this happening, audiotape machines intended for postproduction will have methods of controlling and monitoring spooling speeds when under the control of a synchronizer.

Since a chase synchronizer can only follow the master, both master and slave will have to roll back to a point several seconds before an edit point, as the master has first to reach a stable play speed before the slave machine can lock up to it. It is important that timecode is contiguous during this time, otherwise the slave machine will not be able to lock, and the edit will abort. If necessary, the timecode may have to be restriped if a section of a videotape used as master in post-production has a rogue section of timecode.

The synchronization procedure for the slave machine via a chase synchronizer has three stages: first, the slave cues up to approximately the same position as the master, using timecode as the reference; next, as the master goes into play, the slave machine also goes into play, but with large speed variations possible (up to +50% or so) until the timecode error between the two is less than 1 timecode frame; finally, small speed variations (usually less than ±1%) are used to bring the slave into synchronism with the master, often referenced to individual timecode bits. The slave now locks to the master. Figure 9.4a illustrates.

Control synchronizers

Chase synchronizers are ideal when only one slave machine is required to follow a master. However, when a multi-machine edit facility is required it can be advantageous to employ central control, with a keyboard for addressing individual machines via individual synchronizers, each one programmed with the machine's ballistics. Such a system is referred to as 'modular'. In this way a cue-time address, together with any offsets, can be entered into master and slaves, and all machines, master and slaves, go to their respective addresses individually, in their own time. This will

Figure 9.4 A chase synchronizer (a) will accelerate the slave machine, following the master. A control synchronizer (b) will park the slave at an appropriate point, according to its ballistics, and accelerate both together, handing lock over to an external reference.

reduce cueing time, and allow a system to grow as the requirements expected of it develop.

'Intelligent synchronizers will park machines at a pre-roll point determined for each machine individually, and reduced to a time suitable for the machine with the longest lock-up time. Some synchronizers may park machines with a short pre-roll time, automatically lengthening it incrementally if one machine cannot achieve lock until success is achieved. This facility can be useful where a variety of machines are expected to lock up, and the synchronizer does not recognize a particular machine. This situation can arise when a machine is brought in from outside.

Synchronizers may employ a compromise between locking slaves to the master, and locking everything to external syncs. Known as auto-lock, this system first locks the slaves to the master, to bit accuracy. Control of both master and slaves is then passed to an external reference (Figure 9.4b). This obviates the possibilities of wow and flutter being introduced to the slave transport system from the master. When using auto-lock it is important that the master (usually the videotape recorder) and slave are locked to the same sync source. If a machine is on internal sync it may drift out of sync with the others and the edit will fail.

R-DAT machines used for lay-offs within a video post-production suite will achieve synchronization faster if they are fed with reference syncs (genlocked).

The ESbus

Communications between the various elements of a post-production suite have traditionally been specific to individual manufacturers, with the user unable to transfer such data as EDLs between different editing systems. However, the EBU and SMPTE have agreed a common standard for communication, known as the 'ESbus'. This standard specifies the manner in which individual elements are connected together, signal levels, connectors to be used, data rate and protocol. The complete specification is extremely comprehensive, and is laid down in EBU TECH-3245 and SMPTE 207M, and their various supplements which deal with the application of the ESbus to videotape machines, audio recorders and telecines. The ESbus system is based on the concepts of 'distributed intelligence' and 'virtual machines'. With distributed intelligence, individual devices interpret and act on the information received, their individual controllers attending to such matters as machine ballistics and locking to syncs. Timecode is not sent as a part of the data stream, but as a time address sent to a machine's controller for action. All machines are referenced to a common sync signal called 'Timeline'. This may be derived from a common external source, or be generated internally by a machine synced to a common reference. In this way, messages sent through a system can be kept to a minimum. A virtual machine is an individual machine addressed via its associated controller. This may be a remote-control panel or a synchronizer.

The interface between a virtual machine and the system is called a 'tributary'. The tributary transfers messages to and from virtual machines according to a fixed protocol. Each virtual machine will act on appropriate messages, received from other tributaries, the whole being regulated by a 'bus controller'. Figure 9.5 illustrates.

Figure 9.5 The ESbus system controls individual machine via interfaces (tributaries). According to system needs, common timecode, sync, MTC etc. will be required as ESbus communication data do not carry these on a continuous basis, but as individual 'go-to' or 'event' addresses.

ESbus messages

There are two types of ESbus message, 'virtual machine' and 'system service'. Virtual machine messages pass data between individual virtual machines, and result in either action or response (reply) from the device called. System service messages control the whole system. Data are transmitted asynchronously in serial form at 38.4 kbaud, using RS422 interconnection standards. Each message consists of a series of 8-bit bytes, each preceded by a start bit and followed by a stop bit. The LSB is transmitted first. The connectors used are 9-pin D-type. If twisted-pair telephone cable is used, terminated in 100 Ω, a 1200 m cable run should be possible, though the runs should be kept as short as possible. Typical messages will be Time data, Orders (commands) and Status. Time data messages will be used as a reference by the virtual machine controller section, and will consist of standard timecode addresses. Offsets may be sent, and for audio machines these may contain trimming information down to 1/100 frame. User bit data may also be sent. The messages are not sent continuously, but in response to a command, which will be actioned when the timeline address corresponds to the time data message previously received.

Commands will be used to control a machine, which will respond with a status message by way of acknowledgement. The machine controller will attend to the detail of implementing the command and issuing the status message. Orders include such commands as go to, play timecode source, track select etc. Whenever an order is sent a response is required. Status information is sent by a virtual machine's controller to the system in response to an order. It indicates the activity of the device (stopped, record enabled, playing etc.).

The big advantage of the ESbus over manufacturer-specific systems is that only the controllers associated with individual machines need to be ESbus-capable. This means that, for instance, an ESbus synchronizer can be used with a variety of machines as long as it will interface. Synchronizer manufacturers usually provide machine-specific interface software to cover a wide range of machines at the time of purchase. Updates are available at reasonable cost, so a system may grow with relative ease. Figures 9.6 and 9.7 illustrate a typical small ESbus audio post-production installation. The top synchronizer is the bus controller, the tributaries are interfacing with R-DAT and multi-track machines.

Synchronizer features

It is essential that a synchronizer for audio post-production should allow a tape to be slipped out of sync by a controlled amount, not just by an integral number of timecode frames, but by timecode bit increments. Synchronizers should also have the option of hard or soft lock. Hard lock will follow the master timecode precisely; soft lock will 'flywheel' over

Figure 9.6 Audio Kinetics ESbus synchronizers. The top module is the master.

Figure 9.7 ESbus synchronizers controlling multi-track and R-DAT machines from a VCR master during audio post-production.

any momentary loss of code and timecode jitter. It can be useful to alter the slew rate in soft lock to permit the best compromise between picture/ sound synchronization and minimum audio disturbance. For example, lip-sync requires sync accuracy of better than a frame; but speech is quite tolerant of small degrees of wow. Music, on the other hand may be more tolerant of small picture/sound sync displacements, but is not tolerant of wow.

It can be useful to trim the offset value of timecode down to bit accuracy, particularly 'on the fly' (i.e. dynamically), and for this fresh value to be loaded automatically into the synchronizer for recalculation of the 'in' and 'out' points. The term 'offset' commonly refers to whole (integer) timecode frame shifts, whereas the term 'trim' commonly refers to shift of a fraction of a frame. All synchronizers should be able to read tachometer pulses if any serious audio post-production is to be undertaken. They should also be capable of preventing tape run-off at the end of a reel during spooling. This latter facility should not be needed, but even in the best-regulated world there will still be occasions when timecode has not been recorded contiguously on the master tape, and the slave machines may try to spool through, counting tacho pulses. In this respect, if there is a hole in the timecode addresses the slave machine will have to be taken past the point manually. Unless you are particularly unlucky there will be sufficient contiguous pre-roll timecode the other side of the hole. If not, re-stripe and enter in the new offsets.

An automatic cycling facility will allow an edit to be rehearsed using a play-rewind-play sequence between edit points, or rather between pre-and post-roll points. Coupled with dynamic offsetting and trimming of the slave machine(s), this can be a great help in speeding up the editing process.

Synchronizers can also be set with pre-determined timecode limits. This will prevent a tape spooling off unexpectedly.

Synchronizer problems

If a tape spools unexpectedly, if it spools in the wrong direction, if the play speed varies unexpectedly, you should suspect one of the following causes:

1. No timecode present
2. Gaps in the recorded timecode
3. Timecode is not contiguous
4. The synchonizer is not set for the machine concerned
5. The machine is not under the control of the synchronizer
6. Timecode addresses on the master are not ascending
7. Synchronizers and video machine set at different frame rates
8. The play speed and timecode speed of the slave machine are different (e.g. slave playing at 38 cm/s; timecode striped at 19 cm/s)
9. Synchronizer not programmed with correct tacho pulses data
10. Machine not playing at correct speed for tacho pulses loaded into synchronizer
11. Unwanted offsets in the slave synchronizer(s)
12. Time address limits (if available) not set
13. Different VITC and LTC time addresses on the working copy of the video edit master
14. The R-DAT player is set to a different clock rate from that used to record the tape

If timecode is present but not being read correctly check the following:

1. Dirty timecode heads
2. Unregerierated code coming off tape
3. Timecode at incorrect frame rate
4. Timecode low in level
5. Timecode corrupted

On this last point, timecode is usually recorded on track 24 of an audio multi-track machine. Do not route it through the desk. Opening the wrong fader may be disastrous; there is the possibility of the output crosstalking onto audio, and of audio crosstalking to it, and audio may be inadvertently routed to the same bus as timecode, corrupting it. If an edit keeps aborting, check the following:

1. Pre-roll times are long enough for the slowest slave to lock-up
2. Synchronizers and master videotape machine are all fed with the same reference
3. Master videotape machine set to external syncs
4. Contiguous timecode present during pre-roll
5. Contiguous timecode present during the edit (if the synchronizers require this)

DAT in digital post-production

As we saw in Chapter 4, DAT machines handle incoming (and outgoing) timecode by conversion via lookup tables. As a consequence they are extremely versatile, being able to accept just about any known frame rate. Unfortunately this degree of flexibility can have its disadvantages in postproduction as the timecode off tape behaves as if it were recorded on a longitudinal analogue track, but in fact has been recorded digitally, with all the implications of having been organized into packets and been referenced to either the record DAT's internal clock or to some external clock or sync feed. Problems can arise if the record DAT was recording timecode off the camcorder and not synced to corresponding video. As Figure 9.8 illustrates, the synchronizer in the dubbing suite will try to keep timecode off tape running synchronously with the master timecode by varying the play DAT's speed, whereas the DAT play speed is governed by the wordclock generator or SPG that is the reference for all equipment in the suite. If, at the time of recording, timecode was not synchronous with video the DAT player has a dilemma in that it cannot run its capstan at two different speeds simultaneously.

Figure 9.8 In post-production the digital player's servo system receives sync information from both external timecode and wordsync. If the two sources are not synchronous problems will arise since the machine cannot run at two different speeds simultaneously.

There are ways round this problem, but each has its own shortcomings. The play DAT can be hardlocked to the timecode from the synchronizer, but this leads to distortion, especially on steady tones, as the audio sampling rates between machine and desk differ. The play DAT could be hardlocked to wordclock until the timecode has drifted by a frame, when the DAT's synchronizer will bring it back in sync with the video, accompanied by severe corruption to the soundtrack (blasts of noise). Auto lock (Chase-off) could be used, whereby the DAT machine will chase and lock to the synchronizer then free run at its own speed, with the timecodes drifting slowly apart. As a last resort the analogue outputs of the DAT could be digitally copied by feeding them through the analogue inputs of the desk. There will be a generation loss of quality but this is unlikely to be noticeable.

Timecode and stereo pairs

With a 24-track audiotape machine it is possible to record 10 stereo pairs (track 24 is timecode, track 23 is guardband, track 1 is unused edgetrack, track 2 is spare). When more stereo pairs are required another 24-track machine can be run in synchronism, both machines being locked together by separate synchronizers. Chase mode synchronizers are not suitable for this, as they will try to follow any speed variations in the master. Sets of stereo pairs having a significant proportion of common phase-coherent information should not be placed on separate machines. The machines are best locked by some form of control sync process, so that, once synchronized, they lock to a common and stable reference. Any offset introduced into the synchronizer of one machine should also be entered into those controlling the others.